The Virtual Epileptic Patient (VEP) is a cutting-edge, non-invasive technique in computational neuroscience aimed at creating a digital twin of the brain of a patient suffering from drug-resistant epilepsy and requiring surgical intervention. The VEP enables clinicians to work on a 3D model of the brain, open it up, visualize activity, simulate epileptic seizures and test different approaches or hypotheses as if working on the actual patient’s brain. A clinical study has just concluded with the goal of increasing the success rate of surgical interventions on these patients.
Approximately 1% of the global population is diagnosed with epilepsy, with about 400,000 new cases in Europe each year—or one every minute. Among these individuals, 30% are resistant to medication and necessitate surgical treatment.
Surgery for drug-resistant patients involves destroying the area where seizures originate. This entails making a hole in the skull, implanting one or more electrodes in this area, and then performing thermal destruction of the lesion using these electrodes.
Professor Viktor Jirsa is the scientific lead of the VEP project, Director of the Institut de Neurosciences des Systèmes at Aix-Marseille-Université in the South of France, and Chief Science Officer of EBRAINS. He explained:
“Currently, the success rate for surgery on drug-resistant epileptic patients ranges between 30 and 65%. So, on average, 50% of surgeries are successful, but to put it another way, it also means that 50% of surgeries fail. Thus, there is ample room for improvement.
Surgery failures are typically attributed to the fact that not all epileptogenic areas of the brain have been properly identified. Indeed, it is impossible to place electrodes everywhere into the skull because it is invasive, so surgeons try to minimize the number of electrodes. But this also means that we do not have access to information from everywhere. This is something we can change thanks to the digital twin.”
What is a Digital Twin?
Viktor Jirsa continued:
“A digital twin is a digital representation of an existing person. To create the digital twin of a patient’s brain, we have put in place a methodological capacity to record the data from that person’s brain and input it into the reference space we have designed in EBRAINS at the European level.”
(Credit: EBRAINS)
EBRAINS is a digital neuroscience research infrastructure, launched in 2023 at the European level as one of the key deliverables of the Human Brain Project. It is dedicated to computational neuroscience and the development of digital twins. It aggregates many European hardware resources (included super computers based in France, Germany, Spain, Italy and Switzerland), models and data. Viktor Jirsa said:
“The multilevel brain atlas is crucial because it incorporates in-vivo and ex-vivo brain data from many European laboratories, ranging across many scales of resolution from micrometer to several tens of centimeter (data that are very difficult to obtain), as well as chemistry and time resolved imaging data (such as electroencephalogram (EEG) or functional magnetic resonance imaging (fMRI).
All these data from across European countries have been gathered, made compatible and assembled into a large multimodal atlas, which serves as our reference space. The models are also represented in the same space and made compatible with the data. Our initial experiments to create a digital twin, including our clinical trial, were in the field of epilepsy.”
Entering the Virtual Brain
The Virtual Epileptic Patient models the brain and seizures of patients with drug-resistant epilepsy. These patients require surgical procedures for therapeutic intervention. To reconstruct the patient’s digital brain twin, the team uses recordings such as MRI, electroencephalogram (EEG), magnetoencephalogram (MEG) and stereoelectroencephalography (SEEG).
Creating a digital twin takes approximately 48 hours for virtualization (not counting all the preliminary scans). Viktor Jirsa commented:
“The digital twin of a patient’s brain allows us to visualize a 3D model of the brain, enter it and observe its activity. We are in the process of developing technology that, with a virtual reality headset, allows us to literally plunge our hands into the virtual brain to rotate and manipulate it as desired.
On these 3D models, we can observe the brain in its normal state, when it sleeps or during a seizure. We can also generate and simulate seizure propagation—seizures are not static, but rather dynamic like a wave that begins at a certain point and progresses like a musical score. Thus, we can work on this digital twin as if it were the actual person’s brain but without needing the person.
This is the great advantage of a digital twin: a person can only be operated on once on their brain, and it’s not reversible. With the digital twin, we can operate on it a thousand times, test different approaches and hypotheses, and see the brain’s reactions in real-time.”
Personalization Is Crucial
The effectiveness of the VEP software prototype in providing informative neurosurgical strategies was evaluated from 2019 to 2023 in a European clinical trial involving 356 drug-resistant epileptic patients. This figure is huge considering it is a clinical study in surgery. Viktor Jirsa explained:
“These were prospective patients who were unresponsive to drug treatments and were ready to undergo surgical intervention. We collected imaging information from these patients, created a digital twin for each of them, conducted our analysis by identifying the epileptogenic zone in each of them, and thus the target for surgery. Then we provided this information to the clinicians. They considered our information during their clinical decision making on where and how to surgically intervene on the zone.
The clinical trial has concluded regarding patient inclusion and treatment. However, an epileptic patient is considered seizure free if they do not experience seizures for a year after surgery. So, we are still awaiting this before the final data and results are analyzed.”
He added:
“Today, when we talk about virtual brains, it often involves software or virtual atlas that assists surgeons in navigation. The VEP goes further because it not only allows navigation in the brain, but also predicts the brain’s responses to treatments, stimulation or surgery, and does so in a personalized way.
Because the biggest problem is that two patients with the same symptoms and history may react differently to treatment or medication—patient A might respond well, while patient B might not. Why? Because there’s immense variability in anatomy and physiology from one person to another, or even in connectivity between areas of the brain from one person to another. This variability plays a huge role and is one of the biggest obstacle to progress we face in neuroscience.
That’s why personalization, thanks to the digital twin, is crucial.”
To further personalize their digital twins and improve their models, Jirsa’s team keeps incorporating more and more brain data into their system. Ultimately, they also aim to add other biological data because it’s already known that certain interactions with other organs, such as those between the stomach and the brain for example, are very important.
The Non-Reproducibility of the Human Brain
The human brain is estimated to contain around 86 billion neurons, forming an extensive interconnected network. Within 1 cubic centimeter of human brain tissue, there are no less than 10 trillion synapses, or nerve connections. Neuronal signals travel along nerve fibers from neuron to neuron, with largely varying speeds from 1m/s up to 120 m/s, or 430 km/h.
To enhance speed, the nerve fibers are covered with an insulating sheath called myelin. If we were to line up all these myelinated fibers end to end, we would get a segment measuring 150,000 to 180,000 km, according to estimates. That’s enough to circumnavigate the Earth four times over…
In this context, reproducing an exact replica of a human brain seems to be an impossible mission. Viktor Jirsa commented:
“Yes, it is extremely complex to entirely reproduce a brain, whether it is physical or virtual. And especially, at what level do we stop? Cells, sub-cells, molecules, gels? In principle, today, it’s not possible to recreate exactly a model of the brain, but I can’t rule out that in the future it may become possible.
Actually, it’s not even necessary to mimic every detail, if we want to understand the functioning at a given macroscopic level.”
Next Step: Schizophrenia
Viktor Jirsa and his team have recently begun collaborating with psychiatrists in the field of schizophrenia to replicate what has been achieved in epilepsy and adapt it for schizophrenia. Two clinical trials are planned with several hundred patients.
Schizophrenia, affecting 1% of the global population, exemplifies the challenges inherent in current therapeutic options, with a significant portion of patients experiencing inadequate responses to conventional antipsychotics. For Viktor Jirsa, this poses a real societal challenge:
“By 2030, mental disorders will be the leading cost factor globally in general medicine, surpassing cancers, cardiovascular problems and others. It’s like with climate change—if we wait without taking action until 2030, it will be unmanageable.”
One Aspect That Cannot Be Analyzed
However, according to him, there is still one aspect that cannot be touched or understood with the digital twin.
“There’s something we can’t do and that’s delve into the realm of cognition, i.e., thoughts. We can observe and study how brain activation changes in the digital twin, but brain activation is not cognition. Obviously there is a link and brain activity is the foundation, but the link is not obvious and research on thought and consciousness is a hot topic and well debated.”
This also opens up a whole new field that touches on other areas, such as sociology, ethics or even spirituality.
